llm-jp-13B-instruct-full-jaster-v1.0: benchmark results
LLM-jp v1 instruction-tuned release developed by NII's LLM-jp project, using GPT-2 architecture and a 2,048-token context window. The named instruction-tuning generation remains distinct from the pretrained checkpoint and other instruct generations. Provider: National Institute of Informatics. Access: Open.
Aggregate Rating 1126 ± 42, rank #10409 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| pfgen-bench - Completion Mode - Truthfulness | 0.8 | Truthfulness Score | 61.8 |
| pfgen-bench - Completion Mode - Score | 0.5 | pfgen Score (mean of three) | 57.5 |
| pfgen-bench - Completion Mode - Helpfulness | 0.13 | Helpfulness Score | 53 |
| pfgen-bench - Completion Mode - Fluency | 0.58 | Fluency Score | 52.4 |
| pfgen-bench - QA Mode - Truthfulness | 0.69 | Truthfulness Score | 37 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.69 | MC2 (%) (0-shot) | 28.1 |
| pfgen-bench - QA Mode - Score | 0.28 | pfgen Score (mean of three) | 10.6 |
| Open LLM Leaderboard v1 - HellaSwag | 44.7 | Normalized accuracy (%) (10-shot) | 8.9 |
| pfgen-bench - QA Mode - Helpfulness | 0.01 | Helpfulness Score | 6.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 27.22 | Normalized accuracy (%) (25-shot) | 6.4 |
| Open LLM Leaderboard v1 - GSM8K | 0 | Accuracy (%) (5-shot) | 4.5 |
| Open LLM Leaderboard v1 - WinoGrande | 50.04 | Accuracy (%) (5-shot) | 2.8 |
Interactive version: theaggregate.ai/model?slug=llm-jp-13b-instruct-full-jaster-v1-0 · How It Works · Data refreshed daily, snapshot 2026-11-08.